Know what’s blowing up on TikTok — before anyone else
Someone quietly uploaded 5.6 billion TikTok videos’ worth of numbers — every caption, hashtag, sound and view count since 2014 — free for anyone to question. Not watch. Question.
You don’t need to know SQL. Below are three ready-made questions — tap one open, paste the line, get the answer in under a minute.
The 30-second play
Open the free SQL Console — no account needed
Pick a question, paste its line:
“What’s winning on TikTok right now?”
Top hashtags by total views, last month.
🧾 Show the paste line
SELECT list_element(hashtags,1) tag, count(*) n, sum(views) v FROM 'hf://datasets/datasocial/tiktok-5.6B-videos/data/2026-09.parquet' WHERE len(hashtags)>0 AND is_ad=0 GROUP BY 1 ORDER BY v DESC LIMIT 10
“How much of TikTok is AI-made?”
Videos carrying TikTok’s own AI label.
🧾 Show the paste line
SELECT count(*) FILTER (WHERE is_ai_generated=1) ai, count(*) total FROM 'hf://datasets/datasocial/tiktok-5.6B-videos/data/2026-09.parquet'
“Where does AI content flood the most?”
AI-labeled videos per country, ranked.
🧾 Show the paste line
SELECT country, count(*) n FROM 'hf://datasets/datasocial/tiktok-5.6B-videos/data/2026-09.parquet' WHERE is_ai_generated=1 GROUP BY 1 ORDER BY n DESC LIMIT 10
Read the answer. Teaser: September alone = 3.4 million AI-labeled videos (3.4% of everything posted) — and nobody published that number anywhere. You just made it up yourself.
The edge worth knowing
Sounds break days before songs do — the count of videos using a sound is the earliest signal on the platform. The dataset’s creator wrote it up: Tracking a Billion TikTok Sounds. Best free read this month.
For a hashtag’s full life story — rise, peak, death, revival — the same console goes back to 2014. No marketing agency sells this depth. Nobody has published on this dump yet: every number you pull is first-mover territory.
Source: huggingface.co/datasocial/tiktok-5.6B-videos · 460 GB if you want it all · free for research & tinkering (CC BY-NC)
Fine print: metadata only — no video files · old rows have gaps · counts are snapshots, not live.
Paste a line. Be the one who knew. ![]()
all 3 queries verified live 14 Nov 2026 · links checked · CC BY-NC honored

!